【作者】 刘纲；

【导师】 黄宗明；

【作者基本信息】 重庆大学 ， 结构工程， 2010， 博士

【摘要】 大型桥梁结构的健康监测和安全状态评估能够对结构的安全隐患提前预警从而防止重大灾害事故的发生,是当前土木工程领域的重点研究内容之一,并逐步在工程实际应用中发挥作用。本文利用桥梁恒载在总荷载中占较大比例,对桥梁结构损伤反应相对敏感的特点,提出从长期静态监测数据中提取恒载效应进行大型桥梁安全状态评估的方法,分别从结构恒载效应信息的提取、反演损伤识别方法和静态测点优化布设方法三个方面进行系统研究,探索大型桥梁健康监测和长期安全状态评估新的理论和方法。论文的主要研究内容和研究结果包括:①基于长期静态监测数据的桥梁结构恒载效应信息提取方法研究根据大型桥梁响应中各种作用效应的重现周期,分析其在长期静态监测信号频域中的分布特征,提出了在测试信号中将活荷载、测试误差、温度和混凝土收缩徐变效应信号分离,从而提取恒载效应变化量信号作为损伤信息的方法,主要包括:将k-均值聚类、离子群优化算法、RBF神经网络与巴特沃斯滤波器相结合,提出了自适应变带宽温差作用效应分离方法,利用日温差效应得出温度预测模型,实现年温差效应的分离;基于ACI209、我国现行规范和B3模型的数学理论模型,建立了混凝土收缩徐变增量预测模型,实现了收缩徐变效应的分离;根据活荷载、测试误差效应在监测信号频域的分布特征,采用低通滤波手段将其分离,从而实现了损伤信息的提取。通过数值仿真信号的分析表明:本文提出的分离方法能有效实现从长期健康监测静态数据中提取损伤信息。②基于恒载效应的反演损伤识别方法研究大型桥梁结构单元数量和待识别参数多,而实桥测点布设数量相对很少,导致现有反演损伤识别方法难于收敛到真实解且计算效率难以满足实际需要。针对这两个问题,建立了基于静态数据的复合反演损伤识别方法,以实现大型桥梁损伤程度和损伤部位的判定。在复合反演损伤识别方法中,基于损伤函数引入高维空间降维技术,将损伤识别分为宏单元和有限元单元两个层次,降低了各层次中反演计算待识别变量的个数;根据桥梁破坏通常仅发生在局部的先验知识,引入基于无变化因子的变异操作方法,降低第二代非支配排序多目标遗传算法(NSGA-2)优化求解时的搜索空间;融入了利用数学简化模型加快反演识别计算速度的进化元模型方法,大大提高了复合反演损伤识别方法的计算效率。石板坡长江复线桥的数值仿真算例表明,复合反演损伤识别方法能够根据实桥布设的测点成功识别有限元单元层次的损伤部位和程度,且可将损伤识别的计算效率提高10倍左右。③静态测点优化布设方法研究根据恒载作用下损伤效应具有局部性的特征并采用损伤敏感度最大化原则,提出了基于恒载效应进行测点优化布设的有效影响域法。同时,将相对误差算子融入该方法,实现对不同静态测试参量类型的优化选取。④安全状态评估软件实桥应用研究基于VC++、Matlab和SQL Server数据库技术,利用本文提出的基于长期健康监测数据提取恒载效应信息的方法以及复合反演损伤识别方法,开发了相应的桥梁结构安全状态评价软件,并应用于重庆石板坡长江复线桥长期健康监测的工程实施中。实桥应用表明,本文提出的方法能成功实现日常运营状态下大型桥梁的安全状态评估。更多还原

【Abstract】 Large-span bridges’health monitoring and safety conditions assessment are used to forecast the structural deterioration and prevent the occurrence of disasters, which are major research and application domains in the fields of civil engineering. Based on the high ratio of dead load to total load in large-span bridges, a new safety conditions evaluation method is developed, which utilizes measurement changes in the long-term static monitoring data. The dissertation systematically explores new theories and methods of long-term safety conditions evaluation in the large-span bridge from the views of damage information extraction, damage detection method and optimal sensors placement.The main contents and results are stated in detail as following:①A novel damage information extraction method for long-term static monitoring dataBased on the time scale character of action effects in the long-term static signals, the distribution characteristics in the frequency of signals are analyzed and a new dead load effect extraction method is proposed, which can separate the temperature effect, live load effect, creep and shrinkage effect and measurement error effect from the monitoring signal. Combining the k-means clustering algorithm, particle swarm optimization algorithm, RBF neural network and Butterworth filter, the adaptive band-width temperature separation method is developed. Based on the mathematic prediction model of ACI209, Chinese related regulations and B3 model, the incremental prediction model is presented to separate the creep and shrinkage effects of concrete. According to the distribution location of the live load and measurement error effect in the signal, low-pass filter technique is used to detach them, thus realizing the extraction of the damage information.The experimental verification result of numerical simulation shows that the novel damage information extraction method can accurately acquire the damage information.②A synthetical damage detection method based on dead load effectThe number of sensors in situ bridge is small compared with the number of finite elements and design variables, so the existing methods of condition evaluation based on model updating can hardly converge to true damage condition and the calculation efficiency of these methods is too low to meet the actual need, which prevents the application of these methods. To solve above problems, a synthetical damage detection method based on dead load signal is developed to detect the degree and the position of the damage part of the Shibanpo Yangtze Bridge in Chongqing.The dimension reduction technique is introduced into this method based on the damage function, and the damage detection is departed into two levels, which are level of marco-cell and finite element cell. The number of design variables to be identified in each level is cut down. Based on the priori knowledge that only several locations will be damaged in the whole structure, the non-change factor is used to modify the non-dominated sorting genetic algorithm II(NSGA-2) in order to reduce the search space of the NSGA-2. The meta-model method which substitutes the finite element model with mathematical model is used and calculation efficiency of NSGA-2 is improved obviously.The experimental verification results of numerical simulation show that the synthetical damage detection method can detect the damage location and level successfully in the condition of sparse sensor location. The calculation efficiency can be improved 10 times in the Shibanpo Yangtze River Bridge in Chongqing.③A novel sensor optimal location approach for static sensorAccording to the fact that damage effect of dead load can only be reflected in part of the whole structure and based on the rule of maximization sensitivity of damage, the effective influence domain method for optimal sensors setup is developed, and the relative measurement error factor is added into the method in order to select different static measurement types.④Application of the software system developed to a real bridgeBased on the VC++, Matlab and SQL Server database technique, using the damage information extraction method for long-term static data and the synthetical damage detection method proposed, a software system for condition assessment to bridge is developed and applied to Shibanpo Yangtze Bridge. The application results show that this software can realize the condition assessment to large-span bridges under operation environment.更多还原